def decrypt(self, cipherText):
upperCipher = cipherText.upper()
def swapIndices(a, b, key):
tempKey = list(key)
tempKey[a], tempKey[b] = tempKey[b], tempKey[a]
return "".join(tempKey)
def sample(swaps):
r = random.random()
start = 0
for swap in swaps:
start += swap[0]
if r <= start: return swap
def gibbs(key):
best_swap = (float('-inf'),"")
last_n = []
for i in xrange(200):
for a in xrange(len(key)):
swaps = []
for b in xrange(len(key)):
temp_key = swapIndices(a, b, key)
temp_score = self.languagemodel.score(util.encrypt(upperCipher, temp_key))
temp_swap = (temp_score, temp_key)
if temp_swap[0] > best_swap[0]: best_swap = temp_swap
swaps.append(temp_swap)
# convert to probabilities
maxSwap = max(swaps, key=operator.itemgetter(0))
swaps = [(math.e ** (swap[0] - maxSwap[0]),swap[1],swap[0]) for swap in swaps]
scoreSum = sum([swap[0] for swap in swaps])
swaps = [(float(swap[0])/scoreSum,swap[1],swap[2]) for swap in swaps]
# sample randomly
selected = sample(swaps)
key = selected[1]
# keep last n swaps
converge_n = 5
last_n.append(best_swap)
last_n = last_n[-converge_n:]
# print best_swap[0], util.encrypt(upperCipher, best_swap[1])
# check for convergence
avgSwap = sum([swap[0] for swap in last_n]) / float(converge_n)
if sum([abs(swap[0] - avgSwap) <= 1 for swap in last_n]) == converge_n:
return best_swap
decrypted = util.encrypt(upperCipher, best_swap[1])
sys.stdout.write('.')
sys.stdout.flush()
return best_swap
best_swap = (float('-inf'),"")
num_best = 0
for i in xrange(self.numIters):
key = util.generateKey()
swap = gibbs(key)
if swap[0] == best_swap[0]: num_best += 1
if swap[0] > best_swap[0]:
best_swap = swap
num_best = 1
sys.stdout.write(str(float(i+1)/self.numIters * 100) + "%")
sys.stdout.flush()
translated = util.encryptCase(cipherText, best_swap[1])
print "\n", num_best, "BEST: ", best_swap[0], translated
return translated, best_swap[1], num_best
def main(argv):
learnfile = "ngrams.txt"
testfile = "europarl-v7.es-en.en"
verbose = False
noise = 0.05
numIterations = 0
minLength = 10
maxLength = 60
def printHelpMessage():
print 'decryptor.py [-i <n-gram file> -t <testfile> -n <noise level>]'
print '-v verbose'
print '-h help'
try:
opts, args = getopt.getopt(argv,"hvi:t:n:")
except getopt.GetoptError:
printHelpMessage()
sys.exit(2)
for opt, arg in opts:
if opt == '-h':
printHelpMessage()
sys.exit()
elif opt in ("-i"): learnfile = arg
elif opt in ("-t"): testfile = arg
elif opt in ("-n"): noise = float(arg)
elif opt in ("-v"): verbose = True
print "Learning..."
sys.stdout.flush()
languagemodel = LanguageModel.LanguageModel(learnfile)
original_text_file = open(testfile, "r")
cipher_solver = solver.Solver(languagemodel)
cipher_baseline = baseline.Baseline()
solver_accuracy = []
baseline_accuracy = []
max_counts = []
for original_text in original_text_file:
if len(original_text) < minLength: continue
if len(original_text) > maxLength: continue
numIterations += 1
encryption_key = util.generateKey()
original_text_noised = util.add_noise(original_text, noise)
cipher_text = util.encryptCase(original_text_noised, encryption_key)
startTime = datetime.datetime.now()
if verbose:
print "============================"
print "Iteration ", numIterations
print "Length ", len(original_text)
print "Start Time", startTime
print "Original Text", original_text
print "Original Text Noised", original_text_noised
print "Key", encryption_key
print "Cipher Text Noised", cipher_text
baseline_text, baseline_decryption_key = cipher_baseline.decrypt(cipher_text)
guess_text, guess_decryption_key, num_guesses = cipher_solver.decrypt(cipher_text)
baseline_score = score_accuracy(encryption_key, baseline_decryption_key, cipher_text, original_text)
baseline_accuracy.append(baseline_score)
solver_score = score_accuracy(encryption_key, guess_decryption_key, cipher_text, original_text)
solver_accuracy.append(solver_score)
max_counts.append(num_guesses)
if verbose:
print "End Time", datetime.datetime.now()
print "Duration", datetime.datetime.now() - startTime
print "Length, Accuracy, Duration,", len(original_text), ',', solver_score, ',', datetime.datetime.now() - startTime
print "Baseline Accuracy: ", baseline_score
print "Average Accuracy of Baseline: ", sum(baseline_accuracy)/len(baseline_accuracy)
print "Solver Accuracy: ", solver_score
print "Average Accuracy of Solver: ", sum(solver_accuracy)/len(solver_accuracy)
print "Reached same thing many times", max_counts
print "Average Accuracy of Baseline: ", sum(baseline_accuracy)/len(baseline_accuracy)
print "Average Accuracy of Solver: ", sum(solver_accuracy)/len(solver_accuracy)
print "Over %d cipher texts" % len(solver_accuracy)
